Alexander Reelsen

Alexander Reelsen

@spinscale.bsky.social

Husband, dad, enjoys working distributed, likes distributed databases & search engines, the JVM, Basketball/Streetball fan, gulps coffee, lives in Emsdetten/Germany, occasionally blogs at https://spinscale.de

265 followers471 following742 posts

Using Owntracks & Bunny Edge Script + Database for location tracking Wanted to roughly track my movements using my phone. Pointed Owntracks to a custom Bunny edge script, that also renders a UI. Costs a few cents this month. Provisioned via terraform. Took an hour to get up and running

Alexander Reelsen reposted

Replying to @rafaelcodes.bsky.social

Running a single module scews the measure, too. A big build, where build time matters, will take minutes. Two seconds of them to warm up, maybe, but you will lose two seconds on each forked call to Javac, jar, Javadoc. This cannot work.

Alexander Reelsen reposted

Replying to @spinscale.bsky.social

Unlikely. With repeated builds, most runtime goes to JVM tooling. If you use Javac via the tools API and keep the VM warm through a daemon, you will outperform Rust by length. Warm Java in itself is hard to beat.

I have been dabbling with Langchain4j a little bit in the evening. Instead of using semantic search, I've been using OpenRouter and a bunch of different models to decide if a query is a keyword query ('iphone 16 pro max'), if a query can be build a better using an LLM...

and the existing schema of attributes ('nike hoodie black xxl') or if going semantic makes sense in combination with a chat based interface for product search ('chair without wheels'), so the user can refine the search step-by-step ('exclude brand xyz' or 'expand to show blue and yellow colors').

rowing-machine - Fast + flexible simulated data generator with configurable themes. Written in Rust, has three scenarios (ecommerce, saas, travel), generates csv, jsonl or parquet data, plus optional compression. github.com/gwenwindflow...

github.comGitHub - gwenwindflower/rowing-machine: 🛶 Fast + flexible simulated data generator with configurable themes🛶 Fast + flexible simulated data generator with configurable themes - gwenwindflower/rowing-machine

Finished: Spec driven development Now specs are written for the agents, but also version controlled. This book shows a way of doing it. Hoped for more infos to integrate in existing workflows. Still need to dig into openspec, spec-kit for that I guess Huge font, 150 pages, 29 EUR for the softcover

Finished: Code to cashflow German book, explains the different steps from permanent employee to freelancer to business owner to investor as a path to financial freedom and autonomy. Very short and informative written, however not sure where the AI part really is, except on the cover :-)

Long read/interview with Tom Lane about 30 years of Postgres. Covers scalability, threads, MVCC, query planning vacuuming and many other things. It's always so good to have folks who have worked years on the same piece of software and share the history. www.snowflake.com/en/blog/engi...

snowflake.com30 Years of Postgres Architecture: Tom Lane InterviewExplore 30 years of Postgres architecture with core committer Tom Lane. Learn about process isolation, MVCC, write-ahead logging, and threads.

You may have read the planetscale TIN vs. ParadeDB blog posts already. First the TIN announcement, then ParadeDB with improvements & showing benchmark flaws. Reminder: It's crazy hard to built fair comparisons when you know one product inside out, but not the other. www.paradedb.com/blog/opening...

paradedb.comPlanetScale Released Text Search and We Have a Lot to Say (Part I)Two BM25 optimizations and benchmark configuration changes inspired by PlanetScale's TIN benchmarks make ParadeDB's text search faster without changing its document identifiers.

That little beef aside with my search engineer hat on: nice to see such a focus on integrating search, showing the absolute necessity of full text search in times of ever growing data and agents running queries.

Luxir - a new search engine From the inventor of Solr, written in C++, not clustered for now, uses gRPC, supports streaming import, vector search & hybrid search. Lots of similar terminology with Lucene/Solr/Elasticsearch, so reading the docs will be quite easy. luxir.org

luxir.orgLuxirFull-spectrum search, built to be the fastest. Luxir is an open source search engine with full-text, vector, and faceted search plus analytics in one request, native code that gets more from every…

Finished: Architecting Agentic Solutions with Claude Should help with the first claude certification, however was released 6 months ago, not sure, how up-to-date it is with regards to that. Covers the most high level getting started parts, but not much more.

As I am spending too much time on the couch right now, I watched the Java documentary. Great to see many known faces who have been strong in the Java community for years. And if you look closely at 1:05:54 you will also spot an Elastic logo 😀 www.youtube.com/watch?v=ZqGS...

youtube.comThe Java Story | The Official DocumentaryFrom its humble beginnings as a project code-named "Oak" at Sun Microsystems to becoming a global standard for enterprise software and billions of devices, Java's journey is one of radical…

Dashdown - Use Markdown & SQL to create dashboards Uses tailwind, allows to create own components, realtime updates, has a dozen connectors. Used CSV solar data to try. Seems that evidence now has cloud lock-in, so that someone else created an open source alternative in python. dashdown.ai